Add InstanceInfo to ActionEvent and ActionAttempt protos

[pvditt]

Why are the changes needed?

We want action events to record which cloud instance (node) an action-attempt's pod actually ran on — e.g. p4d.24xlarge — so it can be shown per attempt and used for GPU/cost attribution. Today nothing captures this at runtime.

What changes were proposed in this pull request?

Proto/plugin-core surface only (no executor or run-service behavior change in this repo):

• New InstanceInfo message: instance type, instance id, accelerator, capacity type (spot/on-demand), zone, node name, plus an attributes map for future use.
• ActionEvent.instance_info = 17 — set by an executor when it knows the node.
• ActionAttempt.instance_info = 15 — read-side surface, filled by whatever run service merges events.
• TaskInfo.NodeName in pluginmachinery — lets single-pod execution paths pass the scheduled node name along, so an executor can resolve node metadata off the hot path.

The flyte2 executor/run service does not populate or merge the field yet — consumers of these protos (Union leaseworker) do.

How was this patch tested?

• buf lint passes; Go bindings regenerated with the pinned buf templates.
• go build + existing test suites for runs/service and flyteplugins pass.
• Exercised end-to-end from the Union leaseworker implementation against a local devbox: events carried the field and it round-tripped through GetActionDetails.

Labels

added

🤖 Generated with Claude Code

https://claude.ai/code/session_01PMNn4hYoYsVZWdUB4Cs9iJ

[Copilot]

Pull request overview

This PR extends the run-definition API surface to capture and propagate per-attempt cloud node/instance metadata (e.g., instance type, zone, node name) so action events/attempts can attribute runtime to the underlying compute instance.

Changes:

• Adds a new InstanceInfo proto message and wires it into ActionEvent and ActionAttempt.
• Regenerates Go protobuf + validation bindings to include the new message/fields.
• Extends pluginmachinery TaskInfo with a NodeName field to carry scheduled node name in single-pod execution paths.

Reviewed changes

Copilot reviewed 2 out of 4 changed files in this pull request and generated no comments.

File	Description
flyteidl2/workflow/run_definition.proto	Adds InstanceInfo message and new instance_info fields on event/attempt messages.
gen/go/flyteidl2/workflow/run_definition.pb.go	Regenerated Go protobuf bindings reflecting InstanceInfo + new fields.
gen/go/flyteidl2/workflow/run_definition.pb.validate.go	Regenerated validation code including InstanceInfo validation stubs and embedded validation calls.
flyteplugins/go/tasks/pluginmachinery/core/phase.go	Adds TaskInfo.NodeName for passing scheduled node name through pluginmachinery.

Files not reviewed (2)

• gen/go/flyteidl2/workflow/run_definition.pb.go: Generated file
• gen/go/flyteidl2/workflow/run_definition.pb.validate.go: Generated file

Suppressed comments (1)

flyteplugins/go/tasks/pluginmachinery/core/phase.go:126

• TaskInfo.String() dereferences t.OccurredAt without a nil check, which can panic if a TaskInfo{} is ever formatted/logged (and there are multiple call sites that construct TaskInfo{} without setting OccurredAt). Consider guarding against nil to keep String() safe.

	NodeName string
}

func (t *TaskInfo) String() string {
	return fmt.Sprintf("Info<@%s>", t.OccurredAt.String())

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

Pull request overview

Copilot reviewed 4 out of 8 changed files in this pull request and generated no new comments.

Files not reviewed (3)

• gen/go/flyteidl2/workflow/run_definition.pb.go: Generated file
• gen/go/flyteidl2/workflow/run_definition.pb.validate.go: Generated file
• gen/python/flyteidl2/workflow/run_definition_pb2.py: Generated file

Suppressed comments (2)

flyteplugins/go/tasks/pluginmachinery/core/phase.go:122

• Adding a new exported field to TaskInfo is a source-breaking API change for any downstream code using positional composite literals (e.g. TaskInfo{a, b, ...}), which will no longer compile. This is inherent to Go structs; consider documenting that TaskInfo must be constructed with keyed fields so future additions don't unexpectedly break consumers.

	// Name of the node the task's (primary) pod was scheduled on, when known.
	// Only populated by single-pod execution paths; empty for multi-pod plugins.
	NodeName string

flyteidl2/workflow/run_definition.proto:398

• The comment says capacity_type is normalized to "spot" or "on-demand", but the schema allows any string and has no validation. To keep producers/consumers interoperable, consider modeling this as an enum (preferred) or adding a buf.validate constraint (e.g. string.in) so unexpected values are rejected early.

  // Capacity type of the node, normalized to "spot" or "on-demand".
  string capacity_type = 4;

[wild-endeavor]

add region and provider which is an enum

[wild-endeavor]

@jeevb @EngHabu should we just copy this https://github.com/unionai/flyte/blob/3a2f9ef33da55a017ad447c96523fcd0d3f5d7ec/flyteidl/protos/flyteidl/core/tasks.proto#L51